Sam Yamshon

Trying DeepSeek R1 to make an SVG for a complex org chart. 5 attempts in, it kept leaving critical parts out. I wrote: "Unfortunately, I think an org chart this complex is beyond your capabilities"

It nailed it on the next attempt, with this rationale.

Convince me this isn't AGI!

My ultimate fall/holiday concoction: one sachet each of Harney & Sons pumpkin, apple spice, gingerbread, and cinnamon teas in my giant NewYork Presbyterian cup. Happy Thanksgiving

For the record I think LLMs are a tool that can be wielded in dramatically positive ways, but I worry that these articles are already slipping through the cracks and if they become ubiquitous, have a chance at further eroding what is already a tenuous trust in published science

I’m an associate editor at a journal, and recently have received several articles that were clearly written by ChatGPT (or another LLM). The first ones were obvious, but they’re getting subtle and clever and I think very soon we’re not going to be able to tell the difference…